Methods of encoding and decoding, encoder and decoder performing the methods

Abstract

Provided is an encoding method according to various example embodiments and an encoder performing the method. The encoding method includes outputting a linear prediction (LP) coefficients bitstream and a residual signal by performing a linear prediction analysis on an input signal, outputting a first latent signal obtained by encoding a periodic component of the residual signal, using a first neural network module, outputting a first bitstream obtained by quantizing the first latent signal, using a quantization module, outputting a second latent signal obtained by encoding an aperiodic component of the residual signal, using the first neural network module, and outputting a second bitstream obtained by quantizing the second latent signal, using the quantization module, wherein the aperiodic component of the residual signal is calculated based on a periodic component of the residual signal decoded from the quantized first latent signal output by de-quantizing the first bitstream.

What is claimed is: 
     
       1. An encoding method comprising:
 outputting LP coefficients bitstream and a residual signal by performing an LP analysis on an input signal; 
 outputting a first latent signal obtained by encoding a periodic component of the residual signal, using a first neural network module; 
 outputting a first bitstream obtained by quantizing the first latent signal, using a quantization module; 
 outputting a second latent signal obtained by encoding an aperiodic component of the residual signal, using the first neural network module; and 
 outputting a second bitstream obtained by quantizing the second latent signal, using the quantization module, 
 wherein the aperiodic component of the residual signal is calculated based on a periodic component of the residual signal decoded from the quantized first latent signal output by de-quantizing the first bitstream. 
 
     
     
       2. The encoding method of  claim 1 , wherein the outputting of the second latent signal comprises:
 outputting the quantized first latent signal by de-quantizing the first bitstream, using a de-quantization module; 
 decoding the periodic component of the residual signal from the quantized first latent signal, using the first neural network module; 
 calculating the aperiodic component of the residual signal using the decoded periodic component of the residual signal and the residual signal; and 
 encoding the aperiodic component of the residual signal, using the first neural network module. 
 
     
     
       3. The encoding method of  claim 1 , wherein the outputting of the residual signal comprises:
 calculating LP coefficients using the input signal; 
 outputting the LP coefficients bitstream by quantizing the LP coefficients; 
 determining the quantized LP coefficients by de-quantizing the LP coefficients bitstream; and 
 calculating the residual signal using the input signal and the quantized LP coefficients. 
 
     
     
       4. The encoding method of  claim 1 , wherein the first neural network module comprises:
 a first neural network block to encode the periodic component of the residual signal; 
 a second neural network block to decode the quantized first latent signal; and 
 a third neural network block to encode the aperiodic component of the residual signal. 
 
     
     
       5. The encoding method of  claim 4 , wherein the first neural network block and the second neural network block comprise recurrent neural networks,
 the third neural network block comprises a feed-forward neural network. 
 
     
     
       6. A decoding method comprising:
 outputting quantized LP coefficients, a quantized first latent signal, and a quantized second latent signal by de-quantizing LP coefficients bitstream, a first bitstream, and a second bitstream; 
 outputting a first residual signal by decoding the quantized first latent signal, using a second neural network module; 
 outputting a second residual signal by decoding the quantized second latent signal using the second neural network module; 
 reconstructing a residual signal using the decoded first residual signal and the decoded second residual signal; and 
 synthesizing an output signal using the reconstructed residual signal and the quantized LP coefficients. 
 
     
     
       7. The decoding method of  claim 6 , wherein the second neural network module comprises:
 a fourth neural network block to decode the quantized first latent signal; and 
 a fifth neural network block to decode the quantized second latent signal. 
 
     
     
       8. The decoding method of  claim 7 , wherein the fourth neural network block comprises a recurrent neural network, and
 the fifth neural network block comprises a feed-forward neural network. 
 
     
     
       9. An encoder comprising:
 a processor, 
 wherein the processor is configured to: 
 output LP coefficients bitstream and a residual signal by performing an LP analysis on an input signal; 
 output a first latent signal obtained by encoding a periodic component of the residual signal, using a first neural network module; 
 output a second latent signal obtained by encoding an aperiodic component of the residual signal, using the first neural network module; and 
 output a first bitstream and a second bitstream obtained by quantizing the first latent signal and the second latent signal, using a quantization module, and 
 the aperiodic component of the residual signal is calculated based on a periodic component of the residual signal decoded from the quantized first latent signal output by de-quantizing the first bitstream. 
 
     
     
       10. The encoder of  claim 9 , wherein the processor is configured to:
 output the quantized first latent signal by de-quantizing the first bitstream, using a de-quantization module; 
 decode the periodic component of the residual signal from the quantized first latent signal, using the first neural network module; and 
 calculate the aperiodic component of the residual signal using the decoded periodic component of the residual signal and the residual signal. 
 
     
     
       11. The encoder of  claim 9 , wherein the processor is configured to:
 calculate LP coefficients using the input signal; 
 output the LP coefficients bitstream by quantizing the LP coefficients; 
 determine the quantized LP coefficients by de-quantizing the LP coefficients bitstream; and 
 calculate the residual signal using the input signal and the quantized LP coefficients. 
 
     
     
       12. The encoder of  claim 9 , wherein the first neural network module comprises:
 a first neural network block to encode the periodic component of the residual signal; 
 a second neural network block to decode the quantized first latent signal; and 
 a third neural network block to encode the aperiodic component of the residual signal. 
 
     
     
       13. The encoder of  claim 12 , wherein the first neural network block and the second neural network block comprise recurrent neural networks,
 the third neural network block comprises a feed-forward neural network. 
 
     
     
       14. A decoder comprising:
 a processor, 
 wherein the processor is configured to: 
 output quantized LP coefficients, a quantized first latent signal, and a quantized second latent signal by de-quantizing LP coefficient bitstream, a first bitstream, and a second bitstream; 
 reconstruct a first residual signal by decoding the quantized first latent signal, using a second neural network module; 
 reconstruct a second residual signal by decoding the quantized second latent signal using the second neural network module; 
 reconstruct a residual signal, using the reconstructed first residual signal and the reconstructed second residual signal; and 
 synthesize an output signal, using the reconstructed residual signal and the quantized LP coefficients.